Introduction

First of all, I'd want to give credit to Johnny Bravo as he is the co-author of this article and spent a lot of time testing and helping to find answers to all the questions which were raised during the test weeks. I have learned a lot from him and he was the person who got me to take this board out of the box and try to maximize the performance.

Second, we'd like to elaborate on the reason why exactly we chose Asrock's 4Coredual-VSTA to use in these comparisons. In fact, this is pretty simple: AGP + C2D = ?. We haven't found a lot of AGP boards which gave us the possibility to use the newer Core2Duo processors, which is the best performing CPU series at the moment. Another option would have been to choose a system based on NF3 and an AMD Opteron. But, those chips are limited as well when it comes down to maximum overclock. A good CPU will do 3.2Ghz and that's possible with Intel C2D's as well.

+ = ?

In this article we will guide you through the various aspects of this board, quirks we encountered, chipset limitations, voltage modifications and much more.

Asrock, a little introduction

ASRock Inc., established in 2002, is an energetic company with the combination of technology and humanity. Devoting efforts to bring customers the innovative and reliable motherboards with the design concept of 3C, "Creative, Considerate, Cost-effective", ASRock has successfully established a well-known leading brand of the best price-performance motherboard in the industry.

Facing the constantly changing technologies of motherboard, ASRock will always keep the vision of the future and develop future-proof products upon our 3C design concept to our customers.

It is the commitment to our customers and products, like the spirit presented in our 2004 maxim:

"Fly to the Future with ASRock!"

Basically, Asrock provides low-end motherboards which give the user all options needed to build a reliable system. Asrock is very well-known for its compatibility motherboards, by which I mean the combo motherboards which features DDR1 and DDR2 or AGP and PCIe.

In spite of (or due to) all these handy features, Asrock is not known for its overclocker-friendly motherboards. This article involves thus an attempt to turn a normal motherboard in a perfect overclocking board, perfect as in the ultimate AGP benchmark platform.

Asrock 4CoreDual-VSTA

First of all, this is not the only board of Asrock which can be used for the ultimate AGP benchmark system. The motherboard has been redesigned and updated a few times now, but the basic features we're interested in, stayed the same.

775Dual-880Pro

775Dual-VSTA

4CoreDual-VSTA

4CoreDual-SATA2 R2.0

4CoreDual-SATA2

As this is not an ordinary review, we will only focus on what's interesting regarding the overclocking part: memory compatibility and graphical interface compatibility.





Via PT800 series

Via has been delivering chipset for almost every socket since ages and, as the chipset are mostly a tad slower, has done that for the P4 series as well. A quick glance at the block diagram learns us that the PT880 chipset was not designed for the Intel C2D series as it seems to be designed for 533/800FSB P4 processors. Asrock seems to use 'overclocked' chipsets to provide compatibility for Core2Duo chips.



I've read through the Via white papers a few times, but did not find anything interesting regarding the limits of the chipset. It seems that we will have to find those ourselves...

Bios:

As we handled the board lay-out, we're going to handle the bios option: restrict us to the things we need in order to tune our system.



Right after you've pressed F2 to enter the bios, you'll find yourself looking at the bios main page.



The temperature monitor isn't that elaborate.



The tweaking center of the board, only the two first pages are really interesting.



First issue when overclocking: the bios only supports up to 340FSB. In addition, it's not possible to change the multiplier if you have an unlocked processor.



DDR1 memory timings



DDR2 memory timings



Most voltages can only be set to low/normal/high. Cpu voltage cannot be changed!



Advanced memory configuration tab. We've not looked into this more thoroughly, but this may be the key to tune the memory to higher speeds.






Advanced host configuration, here you can tune the chipset. Please enable the first option (Pipeline DRQCTL) as it gives noticeably higher memory bandwidth.




Who says Asrock doesn't supply lots of tuning options?

Introduction:

In the spirit of overclocking we are attempting to improve on the original, taking the standard and developing it into something greater. As such the hardware will always have limiting factors in this quest for perfection, none more so more as voltage limitations.

We can consider two ways to improve an overclock, run the silicon colder, increase the voltage the silicon runs at, or both (usually necessary!). As the silicon chips are in essence vast arrays of transistors there are associated losses with these devices switching, this is where the heat comes from and heat is the enemy of overclocking. Increased heat causes a number of problems in microchips including increase track resistance and thermal noise. Understandably increasing the voltage will add to the losses, as will increasing the switching frequency (or overclock); therefore it is always wise to upgrade your motherboard’s cooling when increasing the voltages of the components.

The AsRock 4CoreDual-VSTA is not the most forthcoming motherboard in terms of voltage options. In fact there are only two real adjustments, memory voltage and AGP voltage which have all the adjustment of “low”, “normal” and “high” – hardly precise…

In our pursuit of reaching the overclocking limits of this board a few extra voltage options would be needed:

• CPU Core Voltage
  
• CPU VTT Voltage
  
• Northbridge Voltage
  
• Improved AGP Voltage Control
  
• Improved Memory Voltage Control
  
• Readpoints For All the Listed Voltages Above

So let’s take a look at the board.

CPU Voltage modification

1) CPU Voltage read point

Easiest voltage read point to get is the CPU core voltage, simply look for an empty capacitor pad near the socket area and measure from ground to the positive leg, marked on the board in white.



2) CPU Voltage modification

The CPU voltage is controlled by a ST Microelectronics L6714 4 phase PWM controller (.pdf). Normally with a vmod the idea is to alter the feedback bias resistor network by running a suitably sized variable resistor in parallel with the lower (grounded) resistor. The ST controller however as a rather clever little addition to it, an OFFSET input at pin 23 that can be used to introduce an offset to the feedback voltage resulting in an elevated voltage output. A 20Kohm variable resistor in place of the shunt resistor provides a suitable scaling for the vcore voltage.



3) CPU Voltage Vdroop

Vdroop is a recently much debated value. Previously considered the scourge of CPU overclocking it is now slowly becoming accepted as a necessary evil for longevity in a CPU. The AsRock has as an excessive droop, especially at elevated voltages. While this board is certified for quad core use I’m reluctant to say it’ll overclock a quad core with any great success. My personal experience with an Intel X6800 running at 1.6volts was a voltage droop under load of nearly 60mV. This can also be adjusted by adjustment of the external droop network, an example shown here from a user called gustep12:



Memory voltage modifications

1) Memory voltage read point

Similarly for the Memory voltage we find an empty capacitor pad at the top of the motherboard that will provide a suitable read point.



2) Memory voltage modification

The memory is controlled via a Richtek RT9202 synchronous buck PWM controller (.pdf) where we alter the feedback network directly via pin 6 using a 50Kohm variable resistor to ground.



CPU VTT and AGP/MCH modification

1) Cpu vtt read point

The CPU VTT read point is not so clear, but can be found close to the northbridge where there are two MOSFETs located side by side. The CPU VTT voltage can be read off the drain on the MOSFET as shown.



2) Agp/mch read point

When hunting for the AGP and Northbridge voltages it became apparent that the two were In fact sourced from the same point, a MOSFET located between the AGP and PCI-E x16 slots. An empty capacitor pad here allows for measurement of the AGP/NB voltage.



3) Cpu vtt and agp/mch modifications

Voltage control for the CPU VTT and VAGP/VNB can be found near the top of the board beside the DDR/DDR2 sockets. Again the modification involves altering the feedback network, this time by directly altering the feedback network of the opamp. A 10Kohm variable resistor is used of the CPU VTT while a 20Kohm is more suitable of the VAGP/VNB feedback network.



As the AGP and NB voltages are interconnected we are forced to increase the AGP voltage in order to raise the northbridge voltage, in order to increase our FSB/memory speeds. Normally the AGP voltage is around 1.5volts with a usual limit at about 1.7volts. Those of you who remember overclocking on AGP will be no doubt aware that while increasing the AGP voltage did not bring much in the way of improved overclockability, increasing it beyond 1.8volts did run the risk of killing certain cards. Therefore we have a limitation on the northbridge voltage when using an AGP based graphics card.

Taking voltmods even further

Above mentioned issues highlights one of several limitations of the 4coredual-VSTA motherboards design. Due to cost cutting measures we find that circuits have been cut down to their “bare minimum”. Notice that a lot of the capacitor areas are empty, the AGP and northbridge are supplied from the same simple linear MOSFET power supply circuit, the memory power supply for DDR and DDR2 is sourced for the same circuit – as the JEDEC standard for DDR is 2.5 volts I am curious to know how the AsRock engineers are able to boot the motherboard in such a way that it can tell the difference between the memory types and set the voltage accordingly, or is there a situation where DDR2 memory is supplied momentarily with DDR voltages before being reset? Further testing of this power circuit also showed that it wasn’t particularly suitable for high current requirements either, tending to droop under load when using TCCD and Micron “fatbody” based RAM. Testing with Winbond BH-5 based memory seemed to be more suitable, even at elevated voltages of 3.4-3.5volts under load.

To improve upon these design limitations we are forced to take some seemly drastic measures, and start to replace pre-existing components.

For the CPU voltage the best line of action is to upgrade the capacitors with lower ESR (ebullient series resistance) ones to allow the current to flow with less obstruction. My personal choice is Samxon GC and GD series ultra low ESR capacitors that come with a good reputation behind then. You can also at this stage increase the capacitance of the circuit within reason to aid in current capacity. I choose to replace all the capacitors on my board in am attempt to improve the overclockability of various components.



The AGP/NB supply circuit is the Achilles’ heel of the 4CoreDual-VSTA. The MOSFET itself is rather under rated for its duties and has a very low current limit of around 25A. If you were to increase the output voltage to around 1.75volts and attempt to run an AGP card as well you would soon find that the MOSFET literally burns up taking the board with. If you do want to run at higher voltages on the AGP/NB it is imperative that you use a heatsink on this MOSFET.



While I had removed the surrounding capacitors I decided to unsolder this MOSFET as well and replace it with an upgraded model capable of a far larger current. For added protection I installed a small heatsink on top.



Finally the memory power circuit must be able to provide an adequate voltage scale between 1.8 to nearly 4volts in order to encompass DDR2 and DDR in all its guises. To achieve this I have removed the secondary inductor, effectively disconnecting the existing power circuit and attached my own linear design directly to the board. The simpler design has a high current rating and features a variable output from 1.4volts to 4.3volts. It should be noted however that there is no means to automatically switch the voltage scale between DDR and DDR2 so caution must be exercised.



Further improvements could be brought around by replacing the AGP/NB power circuit with a better off board design, separating the AGP and NB power lines, or improving the CPU power circuit further with upgraded inductors.

Test Setup

Massman's Test Setup
CPU	Intel Core 2 E2160 @ 1.8Ghz
Cooling	Aluminium HSF
Mainboard	Asrock 4CoreDual-VSTA
Memory	2 * 256MB PC3200 Twinmoss BH-5 2 * 512MB PC4800 A-Data TCCD 2 * 512MB PC5400 Corsair FatBody
Other	Tagan 480W Western Digital 320GB SATA HDD

Programs used:

Lavalys Everest 4.20

SetFSB

SuperPi 1.5

Cpu-Z 1.43

Methodology: Problems occurring

How did we compare DDR1 and DDR2? Well, to be honest, it was not easy to find a method that allows us to find the real difference between DDR1 and DDR2 as the only option to overclock consists of increasing the FSB. In addition, we must take into account that the only memory dividers provided by Asrock are 133/166/200 for DDR1 and 266/333 for DDR2, so we lack the possibility to use a 1:1 divider platform with both DDR1 and DDR2.

Another problem is the strap change between 232 and 233FSB. Why do we use the word strap? There is not real strap at 233 as Intel never released a processor with 233 as stock FSB. The strap change is in the chipset itself as there's no option to run 1:1 past 233: the divider is automatically changed to 4:3, which results in a huge performance drop. As we're looking for an alternative for DDR2, we need to get higher, much higher.

Third problem seems to be the limitation of the DDR1 memory overclock. The maximum stable overclock we managed to reach was 245FSB (using both 1:1 and 4:3) and the maximum overall was 255MHz. In order to compete with DDR2, we really need to be able to run 265+ cas2 or 300+ cas2.5, but that seemed close to impossible.

Different cpu/memory combinations:

200MHz 1:1 2-2-2-5

200MHz 4:3 2-2-2-5

200MHz 1:1 2,5-3-3-7

200MHz 1:1 2,5-4-4-10

200MHz 4:3 2,5-3-3-7

266MHz 1:1 3-2-2-5

266MHz 1:1 3-3-3-9

Methodology: Solutions

Problem1: Memory dividers
In order to compare both DDR1 and DDR2 at 1:1 divider, we want to compare the results of a 200MHz FSB CPU and a 266MHz FSB cpu, but we need to take the different characteristics of those CPU’s into account. How can we be sure that the results we get are not biased by the CPU itself? Easy, we use the same CPU: in Massman's case an E2160. There are two possible theoretical options:

1) Hope that we can trick the system by setting 266FSB and 1:1 DDR2 works
2) Carry out a BSEL mod

In this case we got lucky and option 1 worked! (To help people who want to carry out a BSEL mod, check out next page's intermezzo.)

Problem2: Strap change
This is a bigger issue, to be honest, but not entirely impossible to work around. Asrock does not provide a windows overclocking tool, though, thanks to SetFSB, we can overclock in windows. The trick is to boot at 232FSB and SetFSB your way up. We can assure you that the program works perfect!

Problem3: MHz limitation
This is the worst issue we have: as there's no way to clock the memory past 255MHz, it seems that DDR2 wins over DDR1 just because of the MHz. We have contacted Asrock for more information, but as far as we can see, it seems to be a chipset related problem. We're hoping for a solution to pop up fast.

To end with, what about 45nm support? Yes and no, the board boots with a 45nm CPU, but is locked in the lowest multiplier setting, which means 6x. This might sound disappointing; however it does mean that 45nm support can be switched on by updating the CPU microcode in the bios. We asked them if this will happen in the future, we're hoping for a positive answer.

BSEL ??

What in God's name does BSEL stand for? Good question! After reading multiple Intel white papers and asking local voltmod guru, Geoffrey, we now know that BSEL means (host) Bus (speed) Select as in the communication between CPU and PLL clock generator. Depending which pins are held low, or made active, the PLL will build a certain bus clock speed which is directly feed to the CPU.

In any case, by changing the BSEL, you change the default FSB to a higher or lower value. You are in fact changing the strap, mostly done because users want to make a certain chip compatible with a new motherboard. For instance, many people use BSEL mods to make their 133FSB run on a new LGA775 motherboard (which only support FSB200+) or to be able to use more memory dividers when the motherboard has no option to change the strap manually.

Before doing any BSEL modification, make sure that you CPU is capable of running at that FSB at stock voltages as the stock multiplier does NOT change!

Digging in deeper

Let's have a closer look at the Intel white papers, concerning the E6x00 C2D series (.pdf):




Basically, all you need to understand is the table. There are three BSEL 'ports' and two states: High and Low. By switching between high and low, we can change the standard strap to 533, 667 or 800. All other values are "reserved", which means that these do not represent any value. Do not try these reserved values as you might damage you motherboard and your CPU. Via Johnny, we asked Intel about these reserved values, but we received no answer yet.

To get back at the actual BSEL modding, there are two states: high represents voltage and low represents ground, however, it's not that simple. When we're talking about voltage, we mean the Vcc, meaning Voltage Collector to Collector or in human English: the core voltage. The ground in this case is the Vss, Voltage Source to Source. The Vcc voltage can as well be controlled by modifications on the motherboard (or the back of the cpu), but in this case, we don't need this. The way to modify is pretty much the same way as modifying the BSEL, but you have to use the VID[7:0] pins. A table for these processor series can be found on page 17 of the Intel white paper.

Working out the BSEL modifications

Now, how does this work. First of all, you need to locate the BSEL0, BSEL1 and BSEL2 pins, which you can locate by, again, looking in the Intel white papers:



Location:
BSEL0: G29
BSEL1: H30
BSEL2: G30



As you can see, the chip is covered with Vcc and Vss pins, so connecting the BSEL to Vcc or Vss (use a silver inkt pen or maybe solder a switch on you motherboard) should not be a problem. Let's help:

- You want an E6300 to run at 333MHz stock, but it runs at 266MHz stock.
1) BSEL at 266 MHz = L-L-L
2) BSEL at 333 MHz = H-L-L
3) Connect BSEL2 to any Vcc pin

- You want your E4300 to run at 266MHz stock, but it runs at 200MHz stock.
1) BSEL at 200 MHz = L-H-L
2) BSEL at 266 MHz = L-L-L
3) Connect BSEL1 to any Vss pin

Note: Please make sure to check the white paper of your processor type before modifying it. For instance, with my E2160, it would have been impossible to do the BSEL modification (according to the Intel white paper) as all other values, except for the 200MHZ, are reserved. As far as I can see, it might be possible that you could apply the same modifications as the BSEL matches exactly the BSEL for the E6x00 series. However, I did not test did and I don't want you to test it for me. If you do this and succeed, please let me know what you found out.

Update: Apparently all values set for LGA775 processors should be compatible, so one could run his C2D at 400FSB strap without a problem. Just have a look at the other Intel whitepapers or give our forums a visit.

Note #2: Only use straps that are supported by your motherboard. For instance, do not mod to 133FSB if you're using a motherboard that only supports straps of 200FSB and higher.

Introduction

To find out what kind of performance DDR1 and DDR2 delivers on this board, we used Lavalys Everest to measure the bandwidth of every memory configuration. Please not that we cannot give you the actual memory bandwidth as it was not possible to lower the multiplier of my E2160 processor, so the bandwidth raised as well because of the increase of memory frequency, FSB frequency as the CPU frequency.

However, these results are not useless as this board has so little overclocking options that when we use it to bench older AGP cards, we will need to overclock the CPU as well. Taking the advantage of the higher CPU speeds into account and compare the memory bandwidth of a particular memory configuration will still be comparable when talking about DDR1 vs. DDR2 performance.

Memory read bandwidth


(click to view full chart)

In our first set of tests, we can see some interesting facts:

- DDR1 2-2-2-5 outperforms DDR2 3-2-2-5.
- DDR1 2,5-3-3-7 is slightly faster than DDR2 3-2-2-5, however the difference is very small.
- 4:3 divider performance is really bad

We included the functions of the most interesting memory configurations that we will use later on this article.

Memory write bandwidth


(click to view full chart)

Other than the Memory Read graph, we can see how DDR1 and DDR2 scale almost the same when running the 1:1 divider. When overclocking, there will be no noticable difference between any DDR type or memory timings, but in theory DDR1 2-2-2-5 seems to be faster. Explanation:

A function of which the graphical representation is a straight line can be written down as a linear equation in the form of "y=ax+b" in which we can find two terms: one term consisting of a constant multiplied by the first power of a variable and a constant.

If X=0, we find the theoretical bandwidth if the FSB=0, which is the value of the constant term 'b'. In this case, we can see that DDR2 is faster in the beginning as the theoretical bandwidth would be approx. 16MB/s.
Now, the value of 'a' is far more interesting, as it tells us how fast the performance will raise. Every time that 'x' is raised by 1, the 'y' value raises by 'a'. In this case, we can see that DDR1 2-2-2-5 scales better than the two other memory configurations, so theoretically 2-2-2-5 should be the best performing configuration (when regarding the memory read bandwidth).

As you might understand, this really is theoretical overclocking and you cannot measure this in real-life overclocking, however keep this basic math in mind as we're going to use it further on the article.

Memory copy bandwidth


(click to view full chart)

Last bandwidth test confirms that 4:3 ... eeemh ... is not that good (what an understatement!). The difference between the 3 most common memory configurations is once again very, very small.

Introduction

Moving on to the more practical side of this article (after the memory latency figures), we're going to give you some SuperPi 1M results. However, don't fall of your chair when giving a first glance on the graph as we have so much more to tell you about this motherboard and overall stability regarding voltages and memory configurations and AGP versus pci-e. But first, we're going to have a look at the latencies.

Memory latency


(click to view full chart)

This is the most difficult graph to interpret on sight, I must admit. Basically it's the same way of the previous figures, but this time the lower the line, the better the performance. Due to the exponential functions, it's might be a bit confusing, however in this case it's pretty logical: the lowest latencies are the fastest.

Then what exactly does this offer us? See how I added the function equations to the graphs: these will help us on the next page.

SuperPi 1M


(click to view full chart)

First of all, everyone knows SuperPi, right? To be short, SuperPi is a very well known benchmark and widely used among overclockers to test stability and performance. It’s a single threaded application and likes well-tuned memory in order to run faster. There are countless tweaks and guides available regarding this benchmark utility, however we don't need these here.

First thing that pops to your mind must be: "Why so little results?". Exactly! During my tests, I've seen very strange things happen regarding stability and Johnny and me agreed on a few pointers:

- A high vMch is not always better; too high means quick instability. This might be solved by added additional cooling such as a peltier or even phase-change cooling.
- A high vAgp is not healthy for your AGP graphic cards, most likely no boots as consequence
- A high vDdr results in very unstable memory, most likely due to the immature memory voltage PWM area
- The CPU vcore PWM area isn't that good either, when reaching 1.6v, the system will become unstable.

Taking this into consideration, I can tell you some more about the instability issues. The reason why so many frequencies are unstable is a mystery to me, however, the error messages I received all lead to one thing: memory instability. The memory was perfectly stable at those frequencies, so I'd rather blame the memory controller built in the Via PT880 chipset. Via is not known for its highly overclockable products and that in combination with such wide compatibility cannot end well.

All I want to conclude from this graph, and I mean that, is that the actual difference between all frequencies is very small.

Why math?

We all know that most overclocking-knowledge comes from trying things out, but in this case, we're so heavily limited by the motherboard (or chipset) that there's not other way than to use the function equations I added to the graphs in order to see what memory configuration is the best performing.

Memory read bandwidth


(click to view full chart)

Equations:
- Cas2: y = 29,38x - 3,87
- Cas2,5: y = 28,64x + 6,032
- Cas3: y = 28,17x + 3,841

As we've told you already, these equations give us the opportunity to predict scores in theory where we could never test at those frequencies. These formulas are quite powerful and thanks to R² being 0,999, or close to 1, we know that the formula is capable of giving quite a good result.

Looking at the equations, we know that in the end Cas2 should be the faster, but the question remains, when? Let's find out when the straight line of the Cas2 equation crosses both other lines. This happens when, for a certain value of X, the result of the difference of two equetion equals zero. In other words:

Cas2 performance = Cas3 performance only if (29,38x - 3,87) - (28,17x + 3,841) = 0
OR 1,21x - 7,711 = 0
OR x = 3,49

The same way we can calculate when the other lines cross each other.

Another way to look at the equations is to calculate which will be faster taking the limits of the memory configuration into account. Let's idealize the system (we'll be using these values for further comparisons as well):
- Decent BH-5 should be capable of running at least 275MHz Cas2
- Decent TCCD should be capable of running at least 310MHz Cas2,5
- Decent FatBody should be capable of running at least 340MHz cas3

To find out which is faster at those speeds, just change the variable x to the proper value and check out yourself. For instance, if I want to know what bandwidth the BH-5 would do at 275Mhz:

y = 29,38x - 3,87
NOW: x = 275
THUS: Bandwidth = 29,38*275 - 3,87 = 8075,63 MB/s

Conclusion:
- Cas2 is faster than Cas3 from 3,49 FSB
- Cas2 is faster than Cas2,5 from 13,38 FSB
- Cas2,5 is faster than Cas3 from -4,66 FSB

- Cas2 275MHz bandwidth: 8075,63 MB/s
- Cas2,5 310MHz bandwidth: 8884,43 MB/s
- Cas3 340Hz bandwidth: 9581,64 MB/s

Memory write bandwidth


(click to view full chart)

Equations:
- Cas2: y = 18,22x - 13,40
- Cas2,5: y = 18,15x + 15,77
- Cas3: y = 18,07x + 15,99<

Conclusion:
- Cas2 is faster than Cas3 from 195,93 FSB
- Cas2 is faster than Cas2,5 from 416,71 FSB
- Cas2,5 is faster than Cas3 from 2,75 FSB

- Cas2 275MHz bandwidth: 4997,1 MB/s
- Cas2,5 310MHz bandwidth: 5642,27 MB/s
- Cas3 340Hz bandwidth: 6159,79 MB/s

Memory copy bandwidth


(click to view full chart)

Equations:
- Cas2: y = 19,42x - 32,55
- Cas2,5: y = 18,53x + 61,75
- Cas3: y = 18,67x + 46,95

Conclusion :
- Cas2 is faster than Cas3 from 106 FSB
- Cas2 is faster than Cas2,5 from 105,96 FSB
- Cas2,5 is faster than Cas3 from 105,71 FSB

- Cas2 275MHz bandwidth: 5307,95 MB/s
- Cas2,5 310MHz bandwidth: 5806,05 MB/s
- Cas3 340Hz bandwidth: 6394,75 MB/s

Memory latency


(click to view full chart)

Equations:
- Cas2: y = 14676(x^-1)
- Cas2,5: y = 14972(x^-0,99)
- Cas3: y = 14329(x^-0,96)

Of course, these are not the type of equations you're used to work with, but the way of comparing still stays the same. For instance, to find where the Cas2 curves crosses the Cas3 curve:

Cas2 performance = Cas3 performance only if 14676/(x^1) = 14329/(x^0,96)
OR (x^0,96)/(x^1) = 14329/14676
OR x^-0,04 = 14329/14676 ~ 0,976356
OR x = | -0,04th root of 14329/14676 |
OR x = | -0,55 |
OR x = 0,55

Conclusion:
- Cas2 is faster than Cas3 from 0,55 FSB
- Cas2 is faster than Cas2,5 from 7,37 FSB
- Cas2,5 is faster than Cas3 from 0,23 FSB

- Cas2 275MHz latency: 53,37 ns
- Cas2,5 310MHz latency: 51,15 ns
- Cas3 340Hz latency: 53,21 ns

Summing up things

Let's have a look what's the fastest clock per clock:

• Memory read bandwidth: Cas2
  
• Memory write bandwidth: Cas2,5
  
• Memory copy bandwidth: Cas2
  
• Memory latency: Cas2

And now with the maximum (theoretical) overclocks taken into account:

• Memory read bandwidth: Cas3
  
• Memory write bandwidth: Cas3
  
• Memory copy bandwidth: Cas3
  
• Memory latency: Cas2,5

Now, we have to admit, we actually hoped DDR1 would come out as winner. DDR1 is definitely not slower than DDR2, but the support of high overclocks is just too poor to really use it as overclocking memory. When benchmarking 3DMark, you'd better choose for DDR2 memory as memory bandwidth will win over memory latency.

In addition, the motherboard is not really capable of running high overclocks and even with voltage modifications, you'd have to be really careful when overclocking as the PWM area is not tuned for that kind of performance. Maybe with some even more insane modifications, this board would be able to break the 360MHz FSB barrier, however, both of us feel that the chipset is just too limiting.

What now? Well, we're still looking into the problems we've come across and are still trying to find solutions to overcome these problems. If we have further updates on this motherboard, we'll definitely post!

If you liked reading this retro-overclocking article, please let us know!

We know this subject isn’t really that popular, but we absolutely like to get involved in datasheet-reading and thinking out of the box to find ways to bypass odd issues. If we get enough response, we'll write other retro articles as well, regarding overclocking and benchmarking older hardware and less high-end material.



Once again, I'd like to thank Johnny Bravo for his help and Asrock as well for answering my questions. Till the next time.